Prediction Of Antigen And Antibody Complex Structure Based On Molecular Simulation And Its Preliminary Application

The use of antibodies is an effective treatment for many important diseases,including severe infectious diseases,immune system diseases and cancers.With the help of computer to simulate the biological behavior of antibody molecules at the atomic level,and to study the binding mode of antigen and antibody from the point of view of antibody structure,the modification and design of antibody molecules as well as the prediction of related physical and chemical properties have been paid more and more attention.Docking is an important method to simulate and predict the structure of antigen-antibody complexes.Although there are many macromolecular docking algorithms and scoring functions,it is still challenging to screen near-native structures which are close to the real situation in thousands of decoys.The main purpose of this study is to establish a method for dealing with docking result of antigen-antibody complexes and for screening the near-native structures from decoys.Thus,the structure of antigen-antibody complex can be predicted effectively by molecular docking and other molecular simulation methods.In this study,the structures of antigen-antibody complex are collected,and the training set and testing set are constructed by molecular docking and cluster analysis.The calculation of each parameter(or interface descriptor)is realized according to the definition.And every parameter is calculated for all the docking conformations(including the near-native structure and the unreasonable structure)in the training set to get sample data.Taking all parameters as the independent variable of the regression model,the logistic regression model(equation)is obtained by using the training set sample data for independent variable selection and regression analysis.According to the probability values calculated by the regression equation for each docking conformation,the predicted near-native structures can be obtained by sorting and selecting the decoys.The characteristic of this method is to sum up the features of antigen-antibody interface according to specific parameters,so that the selected docking conformation have the greatest possibility of meeting the antigen-antibody interface characteristics.The mathematical model is used for distinguishing the near-native structure and the non-reasonable conformation from decoys and it has a pretty accurate and reliable performance in the testing set test and in the prediction of the antibody-binding pattern of ZMapp(neutralizing MAb 4G7,especially)reported in the literature.Furthermore,the structure prediction method of antigen-antibody complex based on molecular simulation was preliminarily applied to predict the key residues of MAb 2B9 for anthrax lethal factor(LF),and to optimize the antibody.The MAb 2B9 was modeled by Discovery Studio 4.5.The crystal structure of anthrax LF was downloaded from PDB database and processed.After docking,5000 simulated conformations(poses)of antigen-antibody were obtained.The parameter values of the independent variables in the logistic regression equation are calculated for each pose,and the probability value that the conformation is a near-native structure is obtained through the calculation of the regression equation.According to the probability values obtained,the docking results are reordered,and the preceding docking conformation is selected for subsequent analysis.Based on the statistical analysis of selected docking poses,the prediction results of the key residues on the antibody are obtained and verified by biological experiments.The overall accuracy of these selected docking simulated conformations is proved.Based on the overall structural information of these selected docking poses,some single point mutants of MAb 2B9 was designed based on the optimization algorithm of antibody point mutation based on the principle of amino acid pairs preference of antigen-antibody interface.The results showed that the antibody mutants with enhanced affinity were obtained and the accuracy of prediction was high.In this study,an antigen-antibody complex structure prediction method based on molecular docking simulation and logistic regression model was established.In addition,the algorithm of optimizational point mutation for antibody based on the structural information of overall docking conformations and the preference of amino acids pairs on antigen-antibody interface was realized.Related software has been developed.The antigen-antibody complex structure prediction method was used to predict the antibody binding model of Ebola virus GP reported in the literature.The prediction results show that the method has high accuracy.The anti-LF MAb 2B9 was optimized by using the above methods,and the enhanced affinity mutants were obtained.The methods established in this study can improve the accuracy and effectiveness of molecular docking to predict the structures of antigen-antibody complexes,and provide useful guidance for the modification and optimization of antibodies.